Identifying pediatric heart murmurs and distinguishing innocent from pathologic using deep learning.

OBJECTIVE: To develop a deep learning algorithm to perform multi-class classification of normal pediatric heart sounds, innocent murmurs, and pathologic murmurs. METHODS: We prospectively enrolled children under age 18 being evaluated by the Division of Pediatric Cardiology. Parents provided consent for a deidentified recording of their child's heart sounds with a digital stethoscope. Innocent murmurs were validated by a pediatric cardiologist and pathologic murmurs were validated by echocardiogram. To augment our collection of normal heart sounds, we utilized a public database of pediatric heart sound recordings (Oliveira, 2022). We propose two novel approaches for this audio classification task. We train a vision transformer on either Markov transition field or Gramian angular field image representations of the frequency spectrum. We benchmark our results against a ResNet-50 CNN trained on spectrogram images. RESULTS: Our final dataset consisted of 366 normal heart sounds, 175 innocent murmurs, and 216 pathologic murmurs. Innocent murmurs collected include Still's murmur, venous hum, and flow murmurs. Pathologic murmurs included ventricular septal defect, tetralogy of Fallot, aortic regurgitation, aortic stenosis, pulmonary stenosis, mitral regurgitation and stenosis, and tricuspid regurgitation. We find that the Vision Transformer consistently outperforms the ResNet-50 on all three image representations, and that the Gramian angular field is the superior image representation for pediatric heart sounds. We calculated a one-vs-rest multi-class ROC curve for each of the three classes. Our best model achieves an area under the curve (AUC) value of 0.92 ± 0.05, 0.83 ± 0.04, and 0.88 ± 0.04 for identifying normal heart sounds, innocent murmurs, and pathologic murmurs, respectively. CONCLUSION: We present two novel methods for pediatric heart sound classification, which outperforms the current standard of using a convolutional neural network trained on spectrogram images. To our knowledge, we are the first to demonstrate multi-class classification of pediatric murmurs. Multiclass output affords a more explainable and interpretable model, which can facilitate further model improvement in the downstream model development cycle and enhance clinician trust and therefore adoption.

Murmurs other than the early diastolic murmur in aortic dissection.

The purpose of this review is to draw attention to the presence and significance of murmurs other than the murmur of aortic regurgitation, in patients with aortic dissection. For that purpose, a literature search was conducted using Pubmed and Googlescholar. The search terms were "dissecting aneurysm of the aorta", "systolic murmurs", "ejection systolic murmurs", "holosystolic" murmurs, "continuous murmurs", and "Austin-Flint" murmur. Murmurs other than the murmur of aortic regurgitation, which were associated with aortic dissection, fell into the categories of systolic murmurs, some of which were holosystolic, and continuous murmurs, the latter attributable to fistulae between the dissecting aneurysm and the left atrium, right atrium, and the pulmonary artery, respectively. Mid-diastolic murmurs were also identified, and these typically occurred in association with both the systolic and the early diastolic murmurs. Among patients with systolic murmurs clinical features which enhanced the pre-test probability of aortic dissection included back pain, stroke, paraplegia, unilateral absence of pulses, interarm differences in blood pressure, hypertension, shock, bicuspid aortic valve, aortic coarctation, Turner's syndrome, and high D-dimer levels, respectively. In the absence of the murmur of aortic regurgitation timely diagnosis of aortic dissection could be expedited by increased attention to parameters which enhance pretest probability of aortic dissection. That logic would apply even if the only murmurs which were elicited were systolic murmurs.

Murmur grading in humans and animals: past and present.

Cardiac murmurs were first described approximately 200 years ago. Subsequently, various clinicians, starting with Samuel Levine, have proposed grading schemes, depicting intensity, or other murmur characteristics, in an attempt to differentiate pathological and physiological murmurs or different degrees of pathology. In the 1960s, these schemes were adapted by veterinary cardiologists and have been used over the last 50 years. However, the clinical utility of these schemes has only recently been examined in veterinary medicine (and never examined in humans), and these studies suggest that the current, commonly used murmur grading scheme is unnecessarily complex and contains redundant information. A simpler, more intuitive grading scheme might achieve the same desired outcome as the more complex scheme, potentially with less confusion. This review examines the history of murmur grading and proposes a reconsideration of the current grading scheme to improve clinical communication.

Automated Identification of Innocent Still's Murmur in Children.

OBJECTIVE: Still's murmur is the most common innocent heart murmur in children. It is also the most commonly misdiagnosed murmur, resulting in a high number of unnecessary referrals to pediatric cardiologist. The purpose of this study was to develop a computer algorithm for automated identification of Still's murmur that may help reduce unnecessary referrals. METHODS: We first developed an accurate segmentation algorithm to locate the first and the second heart sounds. Once these sounds were identified, we extracted signal features specific to Still's murmur. Subsequently, machine learning-based classifiers, artificial neural network and support vector machine, were used to identify Still's murmur. RESULTS: We evaluated our classifiers using the jackknife method using 87 Still's murmurs and 170 non-Still's murmurs. Our algorithm identified Still's murmur accurately with 84-93% sensitivity and 91-99% specificity. CONCLUSION: We have achieved accurate automated identification of Still's murmur while minimizing false positives. The performance of our algorithm is comparable to the rate of murmur identification by auscultation by pediatric cardiologists. SIGNIFICANCE: To our knowledge, our solution is the first murmur classifier that focuses singularly on Still's murmur. Following further refinement and testing, the presented algorithm could reduce the number of children with Still's murmur referred unnecessarily to pediatric cardiologists.

A novel method for discrimination between innocent and pathological heart murmurs.

This paper presents a novel method for discrimination between innocent and pathological murmurs using the growing time support vector machine (GTSVM). The proposed method is tailored for characterizing innocent murmurs (IM) by putting more emphasis on the early parts of the signal as IMs are often heard in early systolic phase. Individuals with mild to severe aortic stenosis (AS) and IM are the two groups subjected to analysis, taking the normal individuals with no murmur (NM) as the control group. The AS is selected due to the similarity of its murmur to IM, particularly in mild cases. To investigate the effect of the growing time windows, the performance of the GTSVM is compared to that of a conventional support vector machine (SVM), using repeated random sub-sampling method. The mean value of the classification rate/sensitivity is found to be 88%/86% for the GTSVM and 84%/83% for the SVM. The statistical evaluations show that the GTSVM significantly improves performance of the classification as compared to the SVM.

A multi-spot exploration of the topological structures of the reconstructed phase-space for the detection of cardiac murmurs.

Acoustic heart signals are generated by a turbulence effect created when the heart valves snap shut, and therefore carrying significant information of the underlying functionality of the cardiovascular system. In this paper, we present a method for heart murmur classification divided into three major steps: a) features are extracted from the heart sound; b) features are selected using a Backward Feature Selection algorithm; c) signals are classified using a K-nearest neighbor's classifier. A new set of fractal features are proposed, which are based on the distinct signatures of complexity and self-similarity registered on the normal and pathogenic cases. The experimental results show that fractal features are the most capable of describing the non-linear structure and the underlying dynamics of heart sounds among the all feature families tested. The classification results achieved for the mitral auscultation spot (88% of accuracy) are in agreement with the current state of the art methods for heart murmur classification.

A review of signal processing techniques for heart sound analysis in clinical diagnosis.

This paper presents an overview of approaches to analysis of heart sound signals. The paper reviews the milestones in the development of phonocardiogram (PCG) signal analysis. It describes the various stages involved in the analysis of heart sounds and discrete wavelet transform as a preferred method for bio-signal processing. In addition, the gaps that still exist between contemporary methods of signal analysis of heart sounds and their applications for clinical diagnosis is reviewed. A lot of progress has been made but crucial gaps still exist. The findings of this review paper are as follows: there is a lack of consensus in research outputs; inter-patient adaptability of signal processing algorithm is still problematic; the process of clinical validation of analysis techniques was not sufficiently rigorous in most of the reviewed literature; and as such data integrity and measurement are still in doubt, which most of the time led to inaccurate interpretation of results. In addition, the existing diagnostic systems are too complex and expensive. The paper concluded that the ability to correctly acquire, analyse and interpret heart sound signals for improved clinical diagnostic processes has become a priority.

Continuous murmur--the auscultatory expression of a variety of pathological conditions.

Continuous murmur is a peculiarity of cardiovascular auscultation, relatively rare, which often hides complex cardiovascular diseases. This article is a review of literature data related to the continuous murmurs accompanied by commenting and illustrating them through our own cases.Recognizing of a continuous murmur and understanding the cardiovascular pathologies that it can hide, is a challenge in current practice.

Heart murmur classification with feature selection.

Heart sounds entail crucial heart function information. In conditions of heart abnormalities, such as valve dysfunctions and rapid blood flow, additional sounds are heard in regular heart sounds, which can be employed in pathology diagnosis. These additional sounds, or so-called murmurs, show different characteristics with respect to cardiovascular heart diseases, namely heart valve disorders. In this paper, we present a method of heart murmur classification composed by three basic steps: feature extraction, feature selection, and classification using a nonlinear classifier. A new set of 17 features extracted in the time, frequency and in the state space domain is suggested. The features applied for murmur classification are selected using the floating sequential forward method (SFFS). Using this approach, the original set of 17 features is reduced to 10 features. The classification results achieved using the proposed method are compared on a common database with the classification results obtained using the feature sets proposed in two well-known state of the art methods for murmur classification. The achieved results suggest that the proposed method achieves slightly better results using a smaller feature set.

Varón de edad avanzada con soplo sistólico. Propuesta de un esquema de diagnóstico diferencial / Elderly male with systolic murmur. A proposal of a differential diagnosis schema

La presencia de un soplo sistólico a la auscultación cardiaca es un hallazgo frecuente en los pacientes de edad avanzada. El origen de estos soplos, puede ser debido a causas banales sin trascendencia clínica ni pronóstica o pueden ser provocados por patologías graves que requieren tratamiento específico. La prueba de elección para el estudio de un soplo es, en la actualidad, la ecocardiografía döppler. Sin embargo, la auscultación minuciosa junto con la realización de determinadas maniobras auscultatorias y el estudio del pulso carotideo son técnicas fundamentales en la consulta del médico de familia que nos permiten aproximarnos al diagnóstico y seleccionar a los pacientes para la realización de un estudio ecocardiográfico (AU)

The presence of a systolic murmur on cardiac auscultation is a frequent finding in the elderly patient. The origin of these murmurs may be due to insignificant causes without clinical or prognostic importance or may be caused by serious conditions that require specific treatment. The test of choice to study a murmur is currently the Doppler echocardiography. However, careful auscultation together with the performance of some auscultation maneuvers and the study of the carotid pulse are fundamental techniques in the family medical consultation that allow us to approach the diagnosis and select the patient who need an echocardiography study AU)

Innocent heart murmurs.

Innocent heart murmur is a frequent auscultatory finding in children. The diagnosis is essentially clinical, without need for further investigation. However, excluding heart disease can be a difficult task. This review article describes some features of medical history and physical examination that help in this differentiation. The role of diagnostic tests is also examined.

Feature extraction for murmur detection based on support vector regression of time-frequency representations.

This paper presents a nonlinear approach for time-frequency representations (TFR) data analysis, based on a statistical learning methodology - support vector regression (SVR), that being a nonlinear framework, matches recent findings on the underlying dynamics of cardiac mechanic activity and phonocardiographic (PCG) recordings. The proposed methodology aims to model the estimated TFRs, and extract relevant features to perform classification between normal and pathologic PCG recordings (with murmur). Modeling of TFR is done by means of SVR, and the distance between regressions is calculated through dissimilarity measures based on dot product. Finally, a k-nn classifier is used for the classification stage, obtaining a validation performance of 97.85%.

Detection of cardiac pathology: time intervals and spectral analysis.

AIM: To develop an objective diagnostic method that facilitates detection of noncyanotic congenital heart diseases. METHODS: Heart sounds and murmurs were recorded from 60 healthy children and 173 children with noncyanotic congenital heart disease. Time intervals were measured and spectrum of the systolic murmurs analyzed. Stepwise logistic regression analysis was used to distinguish physiological from pathological signals. The receiver operating characteristic (ROC) curve was plotted to show the classification performance of the model and the area under the curve (AUC) was calculated. The probability cut-off points for calculation of sensitivities and specificities were estimated. RESULTS: The distinguishing variables were the interval from the end of the first heart sound (S(1)) and the beginning of the systolic murmur, respiratory variation of the splitting of the second heart sound, intensity of the systolic murmur, and standard deviation of the interval from the end of the S(1) to the maximum intensity of the murmur. The AUC was 0.95, indicating an excellent classification performance of the model. The sensitivity of 95% and specificity of 72% was achieved at a probability cut-off point of 0.45. Significant cardiac defects were correctly classified. CONCLUSION: Interval measurements and spectral analysis can be used to confirm significant noncyanotic congenital heart diseases. Further development of the method is necessary to detect also insignificant heart defects.

Development and evaluation of methods for structured recording of heart murmur findings using SNOMED-CT post-coordination.

OBJECTIVE: This study evaluated an existing SNOMED-CT model for structured recording of heart murmur findings and compared it to a concept-dependent attributes model using content from SNOMED-CT. METHODS: The authors developed a model for recording heart murmur findings as an alternative to SNOMED-CT's use of Interprets and Has interpretation. A micro-nomenclature was then created to support each model using subset and extension mechanisms described for SNOMED-CT. Each micro-nomenclature included a partonomy of cardiac cycle timing values. A mechanism for handling ranges of values was also devised. One hundred clinical heart murmurs were recorded using purpose-built recording software based on both models. RESULTS: Each micro-nomenclature was extended through the addition of the same list of concepts. SNOMED role grouping was required in both models. All 100 clinical murmurs were described using each model. The only major differences between the two models were the number of relationship rows required for storage and the hierarchical assignments of concepts within the micro-nomenclatures. CONCLUSION: The authors were able to capture 100 clinical heart murmurs with both models. Requirements for implementing the two models were virtually identical. In fact, data stored using these models could be easily interconverted. There is no apparent penalty for implementing either approach.